Peter Eisentraut <peter.eisentraut@enterprisedb.com> writes:
>>> In short: if the Cygwin Perl headers redefine PGDLLEXPORT (unlikely)
>>> or somehow #define "__attribute__()" or "visibility()" into no-ops
>>> (perhaps more likely) then we could explain this failure, and that
>>> would also explain why it doesn't fail elsewhere.
> This could be checked by running plperl.c through the preprocessor
> (replace gcc -c plperl.c -o plperl.o by gcc -E plperl.c -o plperl.i) and
> seeing what becomes of those symbols.
Yeah, that was what I was going to suggest: grep the "-E" output for
_PG_init and Pg_magic_func and confirm what their extern declarations
look like.
> If we want to get the buildfarm green again sooner, we could force a
> --export-all-symbols directly.
I'm not hugely upset as long as it's just the one machine failing.
regards, tom lane